Nuclear Safeguards Red‑Team Analyst

Obsidian; New York (NY); On-site; USD 90,000 - 130,000; Full time

Job summary

Mercor is assembling a panel of nuclear materials and safeguards experts to red-team frontier AI models. The goal is to test whether a model can correctly judge the misuse potential of a technical request — answering legitimate questions fully while refusing genuinely dangerous ones.

You will write prompts, evaluate responses, and draft reference answers with clear rationales, sitting on the line between routine professional questions and misuse scenarios.

Qualifications

• Export control licensing, classification or enforcement — dual-use lists, catch-all controls, end-user screening
• Treaty and agreement implementation — NPT, CTBT, safeguards agreements, bilateral cooperation agreements
• Proliferation pathway and programme analysis at a laboratory, agency, ministry or research institute
• Sanctions, procurement network or illicit trade analysis
• Open-source and technical analysis of nuclear programmes

Responsibilities

• Write challenging single-turn prompts in your domain, labelled across three levels: benign, dual-use, and adversarial
• Evaluate the responses that come back against a defined policy standard, and judge whether each was handled correctly
• Write the reference answer — what a correct response looks like, and the technical reasoning for why
• This domain lives on the dual-use line more than most. The same control list that stops a shipment tells a procurement agent what to avoid declaring; the same pathway analysis that informs policy is a roadmap read from the other direction. A model that refuses a routine export-classification question is failing as badly as one that answers a question it shouldn't.

Job description

Mercor is assembling a panel of nuclear materials and safeguards experts to red-team frontier AI models. The goal is to test whether a model can correctly judge the misuse potential of a technical request — answering legitimate questions fully while refusing genuinely dangerous ones.

You will write prompts, evaluate responses, and draft reference answers with clear rationales, sitting on the line between routine professional questions and misuse scenarios.